Journal list menu

Article
Full Access

Loss of functional diversity of ant assemblages in secondary tropical forests

First published: 01 March 2010
Citations: 80

Corresponding Editor: D. H. Feener, Jr.

Abstract

Secondary forests and plantations increasingly dominate the tropical wooded landscape in place of primary forests. The expected reduction of biodiversity and its impact on ecological functions provided by these secondary forests are of major concern to society and ecologists. The potential effect of biodiversity loss on ecosystem functioning depends largely on the associated loss in the functional diversity of animal and plant assemblages, i.e., the degree of functional redundancy among species. However, the relationship between species and functional diversity is still poorly documented for most ecosystems. Here, we analyze how changes in the species diversity of ground‐foraging ant assemblages translate into changes of functional diversity along a successional gradient of secondary forests in the Atlantic Forest of Brazil. Our analysis uses continuous measures of functional diversity and is based on four functional traits related to resource use of ants: body size, relative eye size, relative leg length, and trophic position. We find a strong relationship between species and functional diversity, independent of the functional traits used, with no evidence for saturation in this relationship. Recovery of species richness and diversity of ant assemblages in tropical secondary forests was accompanied by a proportional increase of functional richness and diversity of assemblages. Moreover, our results indicate that the increase in functional diversity along the successional gradient of secondary forests is primarily driven by rare species, which are functionally unique. The observed loss of both species and functional diversity in secondary forests offers no reason to believe that the ecological functions provided by secondary forests are buffered against species loss through functional redundancy.

Introduction

Current rates of extinction are estimated to be 100–1000 times greater than rates estimated for pre‐human periods (Lawton and May 1995, Pimm et al. 1995). In particular, tropical forests are likely to experience a large reduction in biodiversity should current trends in human activity continue. A major pressure on biodiversity is the destruction of primary tropical forests and their conversion into secondary habitats. The magnitude of loss in biodiversity depends on the ability of these secondary habitats to act as refuges for forest‐adapted species. The changes in species diversity due to destruction and conversion of tropical forest are relatively well documented, with most papers reporting a reduction in species diversity. Much less is known about the potential effects of species loss on ecosystem functions and services.

Experiments on relatively species‐poor assemblages indicate a positive relationship between species diversity and ecological functions (Naeem et al. 1995, McGrady‐Steed et al. 1997, Tilman et al. 1997). However, this link is neither strong nor universal (Díaz and Cabido 2001, Hooper et al. 2005, Petchey and Gaston 2006). Functional diversity has been defined as “the value and range of those species and organismal traits that influence ecosystem functioning” (Tilman 2001:109). It is emerging as an important aspect of biodiversity as it determines the strength and shape of the relationship between species diversity and ecosystem functions (Díaz and Cabido 2001). The degree to which species perform similar ecological functions in communities and ecosystems, i.e., the level of functional redundancy, is especially important for this relationship (Walker 1992, Lawton and Brown 1993). For example, if all species have an equal and additive effect on function (i.e., functional redundancy is low), one might expect a linear relationship between species diversity and the rate of ecosystem processes. If, on the other hand, many species are functionally redundant, the relationship between species diversity and ecosystem processes should become curvilinear. However, the relationship between species diversity and functional diversity in species‐rich, natural assemblages is poorly understood (Naeem 2002). To achieve the long‐term goal of restoring and managing sustainable ecosystems it is important to understand the linkages and mechanisms between species diversity and ecosystem processes, rather than focusing on species diversity as such (Walker 1992). High functional redundancy of species assemblages might indicate that ecosystem functions are robust to changes in diversity. This has important implications for the conservation of biodiversity and ecosystem functions in (regenerating) tropical forests.

A number of methods have been proposed for measuring functional diversity and richness (Tilman 2001, Mason et al. 2005, Petchey and Gaston 2006, Walker et al. 2008). Most ecological research has relied on the number of functional or trophic groups as a measure of functional diversity, though such approaches have disadvantages (Petchey and Gaston 2006). One disadvantage is the disregard for functional differences within organisms of the same group. More recently, measures of functional diversity have been proposed based on the large functional differences that delineate functional groups, as well as the smaller functional differences within these groups (Petchey and Gaston 2006, Walker et al. 2008). Regardless of the method, all measures of functional diversity suffer limitations. For example, the number and type of functional traits together with their correlations might alter the level of functional redundancy that assemblages appear to exhibit (Fonseca and Ganade 2001, Petchey and Gaston 2002b). Thus, research on the relationship between functional and species diversity must also evaluate the sensitivity of results to the functional traits used.

Here we focus on the functional diversity of ant assemblages (Hymenoptera: Formicidae) along a gradient of secondary succession in the Atlantic Forest of Brazil. From a functional perspective, ants play important roles in terrestrial ecosystems. Firstly, ants are unique because of their ubiquity and abundance in terrestrial ecosystems (Fittkau and Klinge 1973, Hölldobler and Wilson 1990, Tobin 1994). Secondly, ants interact with their environment by performing a variety of ecological functions. These include their functions as seed dispersers (Beattie 1985, Levey and Byrne 1993), predators (Kaspari 1996a, Philpott and Armbrecht 2006), and ecosystem engineers (Lobry de Bruyn and Conacher 1990, Folgarait 1998). In a meta‐analysis on faunal recovery in tropical secondary forest, Dunn (2004) found a general increase in ant diversity along gradients of forest succession (but see also Belshaw and Bolton 1993). It may take, however, several decades for the total recovery of community structure. We have demonstrated a similar pattern for the recovery of ant diversity in our study region in the Atlantic Forest of Brazil (Bihn et al. 2008a). For the same study region, ant behavior at baits indicates an abrupt shift from a preference for protein‐based baits in early successional stages to a preference for carbohydrate‐based baits in late‐successional stages of secondary forests, which might affect the functional composition of ant assemblages (Bihn et al. 2008b). Here we examine the relationship of species diversity to functional diversity of ant assemblages along the same gradient. Specifically, we address the following questions: (1) What is the relationship between species diversity and functional diversity in ant assemblages of tropical forests? (2) How do changes in species diversity along a gradient of regenerating tropical forests affect the functional diversity of ant assemblages?

Methods

Sampling of ants

The study was carried out in the Rio Cachoeira Nature Reserve (25°18′51″ S, 48°41′45″ W) located near the city of Antonina, in the coastal region of the Brazilian state of Paraná. Dense, ombrophilous lowland and submontane forests originally covered the area, but these suffered intense exploitation and large parts of them had been converted to pastures. The resulting landscape mosaic consists of old‐growth forests and secondary forests in various stages of succession (Ferretti and Britez 2006). Between June and September 2003 we sampled leaf litter ants in 12 study sites scattered across the reserve. The sites comprised a chronosequence of four stages of secondary forest succession, with three site replicates for each successional stage: very young secondary forest (4–6 years), young secondary forest (10–15 years), old secondary forest (35–50 years), and old‐growth forests (>100 years). Sites of secondary forests had been used as pastures for buffalo ranching, and site age is given as years after abandonment of ranching. Land use history for study sites was established through interviews with residents and reserve staff corroborated by inspection in a geographic information system (GIS) environment of high‐resolution, geocoded orthophotos from the years 1952, 1980, and 2002. Replicated sites of a particular successional stage were separated by an average distance of 4 km (range = 1–6 km). Replicated sites of a given successional stage were never situated in one continuous patch of the same vegetation type, but separated by areas of different successional stages, pastures, etc. (see Bihn et al. [2008a] for a map of the reserve and the location of the study sites within it).

At each study site we established two 50‐m transects (parallel, separated by 20 m) and collected leaf litter samples (from 1‐m2 quadrats) at 5‐m intervals along these transects (10 sampling points for each transect). This resulted in 20 samples for each site. Transects were located at least 50 m from any trail, pasture, or any other habitat in order to minimize edge effects. Ants were extracted from leaf litter, dead wood, and debris collected from the quadrats by sieving through a 1‐cm mesh screen and subsequently keeping the sifted material in Winkler bags for three days (see Agosti et al. [2000] for a detailed description of the method). All ants were examined and identified by J. H. Bihn. Many ants had to be assigned to morphospecies because they were undescribed or current systematic knowledge is insufficient to assign valid names. For morphospecies mentioned here, the genus name is followed by an epithet in the form “JHB00.” Otherwise, nomenclature follows Bolton (Bolton 1994, 2003). Voucher specimens are deposited at the Museu de Zoologia da Universidade de São Paulo, Brazil (MZUSP), and the State Museum of Natural History Karlsruhe, Germany (SMNK).

Southwood (1996) noted that terrestrial insect assemblages are continually challenged by a flow of transient species (also termed tourist, vagrant, or occasional species). The proportion of transient species is thought to be high in moist tropical forests (Stevens 1989, Longino et al. 2002). For a meaningful analysis one needs to exclude transient species, because these are not biologically associated with the sampled habitat (Magurran and Henderson 2003). Low numbers of individuals and/or low biomass in samples might indicate that a species had not established colonies in the (micro)habitat sampled and probably point to it being transient. Therefore, we excluded from all further analysis species that met at least one of the following criteria: (1) the number of individuals in all combined samples from a site was less than three and (2) the biomass (estimated from the number of individuals and head length with the formula given in Kaspari and Weiser [1999]) of all worker ants from all 20 samples combined was below 0.5 mg (see Appendix A).

The latter criterion is based on the intuitive idea that resource use of a species is proportional to total biomass. Therefore, the exclusion of species with the lowest biomass leads to assemblages dominated by species with a strong impact on resource use and ecosystem functions. The exact limit of 0.5 mg was arbitrary and mainly motivated by the minimum mass needed for the stable isotope analyses. Additionally, we excluded all army ants (Ecitoninae) from our analysis because their occurrence cannot be estimated in a reliable way with the sampling methods employed. We also excluded all males and queens from our analysis because these might never establish colonies after dispersal. This filtering of the original species lists did not qualitatively change the pattern of species richness along the successional gradient (see Appendix B).

Functional traits

Our intention was to assess the functional diversity of ant assemblages with regard to resource use. Four traits were therefore selected that represent (1) the quantity of resources consumed; (2) the mode of resource acquisition; and (3) the type of resources consumed. The functional traits measured for each species were as follows.

Body size.—

We used head length as a measure of total body size because of the strong correlation between head length and body mass (Kaspari and Weiser 1999). Body size is generally considered to be one of the most important attributes of an organism because it correlates strongly with many physiological, ecological, and life‐history traits (Peters 1983). Specifically, the body size of an organism determines the quantity of resources consumed. Head length was measured as the maximum longitudinal length from the most anterior part of the clypeus to the occipital margin, in full face view.

Relative eye size.—

Larger eyes offer a larger visual field and larger visual overlap of the fields. Ant species with large eyes have excellent vision and are very good at detecting moving objects (Via 1977, Wehner et al. 1983), whereas in ants with reduced eyes, optical cues are of minor importance for orientation and foraging. Eye size is also likely to correlate with the main foraging period (diurnal vs. nocturnal). We measured relative eye size as the ratio of eye length to head length.

Relative leg length.—

Longer legs allow faster and more efficient locomotion and foraging (Feener et al. 1988, Franks et al. 1999), but also increase their cross‐sectional area, which could prevent them from utilizing some foraging niches and types of shelter (Kaspari and Weiser 1999). Thus, relative leg length might yield information about the mode of resource acquisition. Relative leg length was measured as the ratio of leg length (combined length of femur and tibia) to head length.

Trophic position.—

Most leaf litter ants in tropical forests are thought to be omnivorous and opportunistic feeders, which harvest plant exudates, scavenge, and capture live prey as these are encountered (Hölldobler and Wilson 1990). However, for the majority of ant species the relative contribution of different food types to their diet is unknown. The analysis of stable isotope composition of organisms provides an alternative approach to assess their trophic position in food webs (Blüthgen et al. 2003, Davidson et al. 2003). A general result obtained from isotope studies is that consumers have relatively higher 15N/14N ratios than their prey. We accordingly used stable isotope data (i.e., 15N/14N ratios) to quantify the trophic position of ants.

For the morphological traits (body size, relative eye length, relative leg length) we measured up to five randomly selected (worker) individuals (average, 3.6; range, 3–5 individuals) and used the mean of these measurements as the value for each species. In species with distinct minor and major worker castes, we only considered minor workers. Each sample for stable isotope analysis typically contained five workers, but often fewer or more depending on size (range, 2–20 workers). For each species we analyzed one sample for every study site (n = 12) in which it occurred, randomly selecting ant individuals from the leaf litter samples collected at that site. Samples were oven dried at 60°C for 48 h, after removal of the gaster to eliminate the effect of undigested food on isotope measurements. Isotopic N composition of each sample was measured using an elemental analyzer–isotope ratio mass spectrometer (EA–IRMS) coupling (EA type 1108, Carlo Erba, Milano, Italy; ConFlo III interface and gas‐IRMS delta S; both Finnigan MAT, Bremen, Germany). The deviation of the sample (smpl) from the international standard (std) in per mil (‰) is expressed as
urn:x-wiley:00129658:media:ecy2010913782:ecy2010913782-math-0001
where Rsmpl denotes the ratio between the heavy isotope and its lighter counterpart (Rsmpl = 15N/14N) for the sample and Rstd denotes the ratio for the international standard (N2 in the air). The N2 from lecture bottles calibrated against the reference substances N1 and N2 for the N isotopes was used as laboratory standard (Gebauer and Schulze 1991). Reference substances were provided by the International Atomic Energy Agency, Vienna. Acetanilide (Merck, Darmstadt, Germany) was used to control the reproducibility and to calibrate N concentration measurements (Gebauer and Schulze 1991).

Species and functional richness and diversity

We calculated functional diversity indices for the ant assemblage in each study site using two widely used indices, following the methods of Petchey and Gaston (functional diversity [FD]; Petchey and Gaston 2002b, 2006) and Walker et al. (functional attribute diversity [FAD]; Walker et al. 1999). Previous studies on the functional composition of ant communities have used the classification of ants into functional groups (e.g., Andersen 1995, 1997). We decided to use continuous measures of functional diversity for two reasons: first, these measures do not require an arbitrary assignment of species into categories (Simberloff and Dayan 1991); second, continuous measures of functional diversity include the functional differences between species within functional groups as well as the differences among functional groups (Petchey and Gaston 2002b).

For the computation of FD, the species by trait matrix was converted into a distance matrix, and this was clustered to produce a dendrogram. We used z‐standardized values to assign all functional traits equal importance in our analysis. The FD of an assemblage is defined as the combined length of all branches in this dendrogram. The choice of distance and clustering method for the calculation of FD may greatly affect the FD values. Thus, we tested several distance and clustering methods (including consensus trees), then selected the most reliable tree for the calculation of FD based on the cophenetic correlation between pairwise distances in trait space and pairwise distances across the dendrogram (see Mouchet et al. [2008] for the distance and clustering algorithms used and details of the method). The combination of Euclidean distances and the unweighted pair group centroid method (UPGMA) yielded the strongest cophenetic correlation (0.85) and were used throughout our analysis. The FAD of an assemblage is the total of all pairwise distances between species in functional trait space. Again, we used Euclidean distances as a measure of dissimilarity.

In their original form, both indices of functional diversity weight every species in a given assemblage equally, i.e., they do not take into account the relative abundance of species. These unweighted indices are therefore measures of functional richness. For exploration of the relationship between functional richness and species richness we used the (unweighted) FD index and the observed species richness per site. For the calculation of functional and species diversity we applied a rarefaction technique to the FD index and species richness. This resulted in indices that account for evenness in species assemblages. The contribution of each species of an assemblage is weighted by its relative number of occurrences per site, i.e., the number of samples (n = 20) per site in which it was recorded.

We plotted values of functional richness and diversity against species richness and diversity and tested for saturation using multiple regression with species richness and quadratic species richness as predictor variables. The analysis was repeated for the relationship between species and functional richness using different combinations of only three of the four functional traits to evaluate whether our results were robust with respect to the number and combination of functional traits considered.

Functional diversity during forest succession

For the comparison of functional diversity indices among ant assemblages we followed the functional rarefaction method as proposed by Walker et al. (2008). As an improvement of Sanders' (1968) rarefaction method, Hurlbert (1971) introduced the idea of using the expected number of species in a sample of n individuals drawn at random from the pool of N individuals as a measure of species diversity. Since each n defines a separate diversity measure, a family of diversity measures can be obtained with the rarefaction technique. For a sample size of n = 2 this measure is equivalent to the Simpson diversity index (Smith and Grassle 1977). Simpson's index is relatively unaffected by rare species. As the rarefied sample size n increases from 2, diversity indices are obtained that are progressively more sensitive to rare species. This property of the indices can be used to understand the contribution of rare and common species to diversity. Walker et al. (2008) generalize the rarefaction technique from species diversity studies and apply it to the functional diversity indices FD and FAD. The effect of rarefied sample size on the sensitivity to rare species is preserved in this generalization of the rarefaction technique. Functional rarefaction transforms the unweighted indices FD and FAD into a family of weighted indices and corrects for sample‐size‐induced bias.

For each ant assemblage we calculated the expected species diversity 〈Sn〉, the rarefied functional attribute diversity index 〈FADn〉, and the rarefied functional diversity index 〈FDn〉 for n = 2 and n = 59 occurrences. The number of species occurrences varied from site to site and 59 occurrences was the ant assemblage with the smallest sample size (range = 59–282). Given that 〈FD2〉 = 2〈FAD2〉 (Walker et al. 2008), we did not calculate 〈FD2〉. Note that we used formulas and algorithms for abundance‐based rarefaction (sensu Walker et al. 2008), which means that we treated each species occurrence as an individual in these calculations.

We made use of the following three properties of the calculated rarefied indices for the interpretation of the results (Walker et al. 2008). First, the functional diversity indices 〈FAD2〉, 〈FAD59〉, and 〈FD59〉 are sensitive to the differences in the functional traits of species whereas the indices of expected species diversity 〈S2〉 and 〈S59〉 are not. Second, the rarefaction to different numbers of occurrences gives way to the evaluation of the sensitivity to rare species: 〈S59〉, 〈FAD59〉, and 〈FD59〉 are more sensitive to rare species than 〈S2〉 and 〈FAD2〉. Third, Petchey and Gaston (2006) emphasize that FD is insensitive to functionally redundant species whereas 〈FAD2〉, 〈FAD59〉, 〈S2〉, and 〈S59〉 are sensitive to these.

Results

Our analysis was based on 2212 occurrences of 99 species (30 genera) in 12 sites. Ant species covered a wide range of values for the functional traits measured. For example, the mean head length of the smallest ant (0.34 mm; Brachymyrmex JHB02) was almost eight times smaller than that of the largest ant (2.67 mm; Odontomachus haematodus). Mean δ15 N values ranged from 1.94 (Acropyga fuhrmanni) to 10.8 (Amblyopone armigera). Studies on insects generally report a δ15 N enrichment of 2–3‰ per trophic level (McCutchan et al. 2003). Thus, ant species in our study covered about three trophic levels. The functional dendrogram describes the functional relationships among ant species (Fig. 1). This plot highlights that a number of sets of species are functionally very similar, so that if they are present in the same local assemblage they will be redundant with respect to one another. Other species had a very uncommon combination of functional traits, e.g., Pheidole lucretii or Amblyopone armigera. These species always increased the functional diversity of assemblages irrespective of which other species were present.

figure image

Summary of the data set. On the right is a representation of the species occurrence matrix with species as rows and sites as columns. Sites are grouped by successional stage and ordered from left to right by increasing time since abandonment of use as pastures. A black square indicates the presence of a species. On the left is the functional relationship among 99 ant species that were sampled at the Rio Cachoeira Nature Reserve, Brazil. The dendrogram was produced by hierarchical clustering with the unweighted pair group method with arithmetic mean (UPGMA) algorithm of the Euclidean distance matrix calculated from the standardized functional traits of species. For morphospecies the genus name is followed by an epithet in the form of “JHB00”; see Methods: Sampling of ants for an explanation of codes.

We found a linear relationship between functional richness and species richness (FD and S) and between functional diversity and species diversity (〈FD59〉 and 〈S59〉; Fig. 2). For both relationships, the effect of squared species richness was not significant in the multiple regression model (richness, P = 0.79; diversity, P = 0.21). This indicates that linear regression is an appropriate way to describe the relationships. There was a close, linear relationship between species and functional richness (r2 = 0.92, P < 0.001; Fig. 2a) and species and functional diversity (r2 = 0.87, P < 0.001; Fig. 2b). These results proved robust to the use of different numbers and combinations of functional traits (Fig. 3). The amount of variation explained by the relationship between species and functional richness (S and FD) was similar in all cases (all traits, r2 = 0.92; without trophic position, r2 = 0.92; without relative eye size, r2 = 0.85; without body size, r2 = 0.92; without relative leg length, r2 = 0.93). The increase in functional richness with increasing species richness was not influenced by the number and combination of functional traits (ANCOVA, P = 0.97).

figure image

Relationship between (a) functional richness and species richness (FD and S) and (b) functional diversity and species diversity (〈FD59〉 and 〈S59〉) across 12 ant assemblages sampled at Rio Cachoeira Nature Reserve, Brazil. Each species is weighted equally in the richness indices FD and S. We applied a rarefaction technique (rarefied sample size = 59) for the calculation of the diversity indices 〈FD59〉 and 〈S59〉. Each species is weighted by its relative number of occurrences per site in the diversity indices. See Methods for further details on the calculation of these indices.

figure image

Relationship between species richness and functional richness (S and FD) for different sets of functional traits. The traits used for the calculation of functional richness were body size, relative leg length, relative eye size, and trophic position. Each point represents the function and species richness for an ant assemblage using all four traits and all combinations of only three of these traits.

We found increasing values for the indices 〈S2〉, 〈S59〉, 〈FAD59〉, and 〈FD59〉 along the successional gradient (ANOVA, linear trend; 〈S2〉, F1,8 = 26.01, P < 0.001; 〈S59〉, F1,8 = 47.29, P < 0.001; 〈FAD59〉, F1,8 = 44.79, P < 0.001; 〈FD59〉, F1,8 = 19.12, P = 0.002; Fig. 4). The index 〈FAD2〉 did not show a clear pattern among the successional stages (ANOVA, linear trend; F1,8 = 0.41, P < 0.54). Because of the complementary properties of the employed indices, we can make several conclusions based on the observed pattern:

figure image

Changes in the functional diversity and species diversity of leaf litter ant assemblages in the Rio Cachoeira Nature Reserve along a gradient of forest succession. Five diversity indices are used: 〈S2〉, 〈S59〉, 〈FD2〉, 〈FAD59〉, and 〈FD59〉. 〈Sn〉 is the expected species diversity after rarefaction, 〈FADn〉 is the rarefied functional attribute diversity index, and 〈FDn〉 is the rarefied functional diversity index. The subscript n indicates the rarefied sample size. See Methods for further details on the calculation of these indices. Circles and error bars are means of estimated levels of diversity with standard errors for each successional stage (n = 3).

  • (1) 〈S2〉 and 〈S59〉 increased with successional age. Thus, species diversity of leaf litter ants increased along the successional gradient.

  • (2) Functional diversity of ant assemblages also increased with increasing successional age, and species that are recruited to assemblages are likely to be unique with respect to their functional traits. Two out of three indices of functional diversity, including 〈FD59〉, which is insensitive to functionally redundant species, increased along the successional gradient.

  • (3) The recruitment of rare species was largely responsible for the increase in functional diversity of ant assemblages. Though some common species became more abundant along the successional gradient (〈S2〉 significantly increased), we did not detect a significant increase in the functional differences among common species (as measured by 〈FAD2〉). This suggests that the common species that became more abundant along the successional gradient were not sufficiently unique with respect to their functional traits to result in significant increase in 〈FAD2〉. Thus, the observed increase of 〈FAD59〉 (which is more sensitive to rare species than 〈FAD2〉) resulted mainly from the recruitment of species that are functionally unique and rare.

If our conclusions are correct, the functional uniqueness of ants should be related to their rarity. To check this conclusion, we measured the functional uniqueness of ant species using the quadratic‐entropy (QE)‐based index proposed by Pavoine et al. (2005). This index is a measure of how distinct or unique a species is in its trait values compared with other species in an assemblage. Functional uniqueness of the ant species was negatively correlated to the number of occurrences in samples (Pearson correlation on log10‐transformed values, r = −0.34, P < 0.001). This result confirms that rare ants (e.g., Amblyopone armigera, Acanthognathus brevicornis, Myrmicocrypta JHB02, and Acropyga fuhrmanni) tend to be more functionally unique than common ants.

Discussion

The relationship between species and functional diversity is still poorly documented for most ecological systems (Naeem 2002). We show here that functional richness and diversity are closely related to species richness and diversity of tropical leaf litter ants. The link between species diversity and functional diversity is strong and positive for the entire range of local species richness, with no evidence for saturation in this relationship. These results point to low levels of functional redundancy among coexisting species of leaf litter ants. Recovery of species richness and diversity of ant assemblages in tropical secondary forests were accompanied by an increased functional richness and diversity of assemblages. Moreover, our results indicated that the increase in functional diversity along the successional gradient of secondary forests is primarily driven by rare species that are functionally unique. The decline in ant species diversity in tropical forests due to human alteration of the environment is likely to result in a proportional decline of the ecological functions performed by this taxon.

We quantified the functional diversity of ant assemblages based on traits for which evidence about their functional significance for species' resource use patterns exists (Kaspari 1996b, Gotelli and Ellison 2002, Blüthgen et al. 2003, Davidson et al. 2003). Weiser and Kaspari (2006) demonstrate that body size, relative leg length, and relative eye size explain most of the variation in ecological morphospace among ant species of the New World and that these traits are linked to the foraging behavior of ants. However, a demonstration of the functional significance of traits does necessarily mean that differences in the functional diversity of ant assemblages have consequences for ecosystem functioning. In addition, resource use may be appropriate for some ecosystem functions but not others. For example, many ant species live in close mutualistic associations with plants. A functional classification of species based on resource use traits might only poorly relate to the functional consequences for these plants, and another classification with a different set of traits (e.g., life history traits) might have more explanatory power for these functions. For traits relating to resource use patterns of ants, our results indicate low levels of functional redundancy among coexisting species. Considering that ants carry out different functions at the same time, including functions for which resource use traits are not appropriate, overall functioning of ants is likely to be even more susceptible to changes in species diversity than the results of our study suggest (Gamfeldt et al. 2008).

The observed level of functional redundancy might depend on the number of traits used for the measurement of functional richness and diversity (Fonseca and Ganade 2001, Petchey and Gaston 2002b). The inclusion of a large number of uncorrelated functional traits will inevitably produce assemblages with low functional redundancy. Thus, traits need to be selected with care. Experimental evidence of their functional significance would be the best approach for their selection (Petchey and Gaston 2006). The rather modest number of traits used in our study compared to other studies on patterns of functional diversity does not seem problematic. Most importantly, even when we repeated our analysis with only three traits, a similar relationship between species and functional richness emerged (Fig. 3). Thus, our conservative approach with respect to the number of included functional traits revealed patterns of low functional redundancy consistent with our results for the full set of traits.

Our results allow predictions to be made about the potential gains and losses in functional diversity associated with community assembly during secondary succession. The analysis suggests that small changes in species diversity can have rather large effects on functional diversity, and possibly ecosystem functioning, within a community. Therefore, the conservation of a large proportion of the ecological functions of communities requires the conservation of a large proportion of the species that make up the community. Moreover, our results suggest that rare species often possess unique combinations of functional traits. Since rare species can make significant contributions to ecosystem functioning (Lyons et al. 2005) and are especially prone to extinction (Gaston 1994), anthropogenic disturbances could lead to rapid loss of ecosystem functioning.

The recovery of species and functional diversity followed similar trajectories through secondary succession (Fig. 4). Both measures of diversity were closely related and increased monotonically with increasing age of secondary forests. Nevertheless, it may take many decades until species and functional diversity reach levels similar to those in primary forests. These results confirm studies on the species diversity of ants along successional gradients in the tropics (reviewed by Dunn 2004). For conservation practice this means that the largest possible range of functional traits and probably also ecosystem functions can only be preserved in primary forests. Our results also suggest that classical measures of biodiversity such as species richness and diversity might be good surrogates for functional richness and diversity of communities. The monitoring of the diversity of invertebrates is often time‐consuming and costly. This is especially true for tropical ecosystems where many invertebrate taxa are hyperdiverse. The assessment of functional diversity of invertebrate assemblages will in most cases involve the measurement of a set of functional traits of species, which puts even more strain on resources of time and money. As long as species diversity explains much of the variation in functional diversity, as suggested by our study, the measurement of functional traits is not necessary. The conservation of local species diversity will result in the conservation of an almost proportional amount of functional diversity. At this time, however, it would be premature to draw final conclusions based on our particular system. Indeed, our study is one of the first to explore the relationship between species and functional diversity for a species‐rich assemblage of invertebrates in the tropics. Based on the results of simulated random extinctions of species in a South American plant community, Fonseca and Ganade (2001) conclude that as much as 75% of the species can be lost before changes in the functional diversity (measured as functional group richness) become evident. This seems to be not the case in our study. Differences between their findings and ours could result from different measures of functional diversity used in the studies (number of functional groups in the study of Fonseca and Ganade [2001]), which make comparison difficult. Concordant with our results, Micheli and Halpern (2005) report a strong positive relationship between species and functional diversity and increased functional diversity coinciding with the recovery of species richness in marine reserves. A similar linear relationship was also found in a study simulating the random extinction of species from five animal and plant assemblages (Petchey and Gaston 2002a). In contrast, Ernst et al. (2006) demonstrate that functional diversity of tropical amphibian communities in secondary forest is lower than in primary forests but that this reduction in functional diversity does not always match patterns of species diversity, indicating a rather loose relationship between these two measures of biodiversity. These contradicting results for different taxa and environments highlight the urgent need for studies on the consequences of human actions on the pattern of functional diversity in natural assemblages.

Further studies should also clarify how changes in the functional diversity of natural assemblages impact ecosystem functioning. Our study does not provide a direct link between species diversity and ecosystem functioning and the observed changes in functional diversity might not proportionally translate into changes of any single ecosystem function. The majority of studies on the effect of biodiversity on ecosystem functioning has used synthetically assembled communities and measured the effect of changes in species composition on single ecosystem process rates or properties. Equating single functions with overall functioning may ignore other important ecosystem processes and can be highly misleading (Rosenfeld 2002, Gamfeldt et al. 2008). Our approach might better reflect the multiple ecological functions species provide in ecosystems and avoid the bias towards detecting ecological redundancy by focusing on a single ecological function.

Management and conservation of tropical forests will require a better understanding of the value of secondary forests for biodiversity conservation. Invertebrates provide essential ecological functions in most ecosystems (Wilson 1987). For most invertebrate groups we lack information about the extent to which secondary forests can preserve even relatively simple aspects of biodiversity such as species richness and diversity. In the long term, the link between ecological functions and species diversity might have very important effects on the complex interactions affecting human well‐being. Our results offer a first insight into the nature of these relationships in species‐rich assemblages of tropical forests. The low functional redundancy among species in our study indicates that species loss due to human alteration of the environment will lead to a severe decline in ecological functions. Therefore, secondary forests might not only harbor much reduced species diversity but also offer significantly reduced ecosystem functions in comparison to undisturbed forests. More studies on the functional diversity of assemblages in secondary forests and their impact on ecological functions are needed to provide guidance for the conservation and restoration of tropical forest communities.

Acknowledgments

We are grateful for the support of the Sociedade de Pesquisa em Vida Selvagem e Educação Ambiental and their staff at Rio Cachoeira Nature Reserve, without which the fieldwork could not have been completed. We thank V. Traxel for providing some of the size measurements in this study. Skillful technical assistance in isotope ratio mass spectrometry by Iris Schmiedinger (BayCEER, University of Bayreuth) is gratefully acknowledged. Constructive criticism from R. R. Dunn and an anonymous reviewer greatly improved this manuscript. This study was supported by a grant from the German Federal Ministry of Education and Research within the framework of the SOLOBIOMA project (01LB0201).

    APPENDIX A

    Rank–biomass diagrams for ants in each of the 12 study sites (Ecological Archives E091‐057‐A1).

    APPENDIX B

    Species richness for each successional stage before and after filtering of species lists (Ecological Archives E091‐057‐A2).